{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/training-compute-optimal-large-language","title":"Training Compute-Optimal Large Language Models","arxiv_id":"2203.15556","date":"2022-03-29","proceeding":null,"authors":["Jordan Hoffmann","Sebastian Borgeaud","Arthur Mensch","Elena Buchatskaya","Trevor Cai","Eliza Rutherford","Diego de Las Casas","Lisa Anne Hendricks","Johannes Welbl","Aidan Clark","Tom Hennigan","Eric Noland","Katie Millican","George van den Driessche","Bogdan Damoc","Aurelia Guy","Simon Osindero","Karen Simonyan","Erich Elsen","Jack W. Rae","Oriol Vinyals","Laurent SIfre"],"abstract":"We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over \\nummodels language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, \\chinchilla, that uses the same compute budget as \\gopher but with 70B parameters and 4$\\times$ more more data. \\chinchilla uniformly and significantly outperforms \\Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that \\chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, \\chinchilla reaches a state-of-the-art average accuracy of 67.5\\% on the MMLU benchmark, greater than a 7\\% improvement over \\gopher.","url_abs":"https://arxiv.org/abs/2203.15556v1","url_pdf":"https://arxiv.org/pdf/2203.15556v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"training-compute-optimal-large-language","repo_url":"https://github.com/karpathy/llama2.c","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"training-compute-optimal-large-language","repo_url":"https://github.com/nkluge-correa/teenytinyllama","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"anachronisms","task_name":"Anachronisms"},{"task_slug":"analogical-similarity","task_name":"Analogical Similarity"},{"task_slug":"analytic-entailment","task_name":"Analytic Entailment"},{"task_slug":"causal-judgment","task_name":"Causal Judgment"},{"task_slug":"common-sense-reasoning","task_name":"Common Sense Reasoning"},{"task_slug":"crash-blossom","task_name":"Crash Blossom"},{"task_slug":"crass-ai","task_name":"Crass AI"},{"task_slug":"dark-humor-detection","task_name":"Dark Humor Detection"},{"task_slug":"date-understanding","task_name":"Date Understanding"},{"task_slug":"disambiguation-qa","task_name":"Disambiguation QA"},{"task_slug":"discourse-marker-prediction","task_name":"Discourse Marker Prediction"},{"task_slug":"empirical-judgments","task_name":"Empirical Judgments"},{"task_slug":"english-proverbs","task_name":"English Proverbs"},{"task_slug":"entailed-polarity","task_name":"Entailed Polarity"},{"task_slug":"epistemic-reasoning","task_name":"Epistemic Reasoning"},{"task_slug":"evaluating-information-essentiality","task_name":"Evaluating Information Essentiality"},{"task_slug":"fantasy-reasoning","task_name":"Fantasy Reasoning"},{"task_slug":"figure-of-speech-detection","task_name":"Figure Of Speech Detection"},{"task_slug":"formal-fallacies-syllogisms-negation","task_name":"Formal Fallacies Syllogisms Negation"},{"task_slug":"gre-reading-comprehension","task_name":"GRE Reading Comprehension"},{"task_slug":"general-knowledge","task_name":"General Knowledge"},{"task_slug":"hellaswag","task_name":"HellaSwag"},{"task_slug":"human-organs-senses-multiple-choice","task_name":"Human Organs Senses Multiple Choice"},{"task_slug":"hyperbaton","task_name":"Hyperbaton"},{"task_slug":"identify-odd-metapor","task_name":"Identify Odd Metapor"},{"task_slug":"implicatures","task_name":"Implicatures"},{"task_slug":"implicit-relations","task_name":"Implicit Relations"},{"task_slug":"intent-recognition","task_name":"Intent Recognition"},{"task_slug":"irony-identification","task_name":"Irony Identification"},{"task_slug":"known-unknowns","task_name":"Known Unknowns"},{"task_slug":"lambada","task_name":"LAMBADA"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"logic-grid-puzzle","task_name":"Logic Grid Puzzle"},{"task_slug":"logical-args","task_name":"Logical Args"},{"task_slug":"logical-fallacy-detection","task_name":"Logical Fallacy Detection"},{"task_slug":"logical-reasoning","task_name":"Logical Reasoning"},{"task_slug":"logical-sequence","task_name":"Logical Sequence"},{"task_slug":"mmlu","task_name":"MMLU"},{"task_slug":"mathematical-induction","task_name":"Mathematical Induction"},{"task_slug":"mathematical-reasoning","task_name":"Mathematical Reasoning"},{"task_slug":"metaphor-boolean","task_name":"Metaphor Boolean"},{"task_slug":"misconceptions","task_name":"Misconceptions"},{"task_slug":"moral-permissibility","task_name":"Moral Permissibility"},{"task_slug":"movie-dialog-same-or-different","task_name":"Movie Dialog Same Or Different"},{"task_slug":"movie-recommendation","task_name":"Movie Recommendation"},{"task_slug":"multi-task-language-understanding","task_name":"Multi-task Language Understanding"},{"task_slug":"multiple-choice-qa","task_name":"Multiple Choice Question Answering (MCQA)"},{"task_slug":"navigate","task_name":"Navigate"},{"task_slug":"nonsense-words-grammar","task_name":"Nonsense Words Grammar"},{"task_slug":"novel-concepts","task_name":"Novel Concepts"},{"task_slug":"odd-one-out","task_name":"Odd One Out"},{"task_slug":"penguins-in-a-table","task_name":"Penguins In A Table"},{"task_slug":"phrase-relatedness","task_name":"Phrase Relatedness"},{"task_slug":"physical-intuition","task_name":"Physical Intuition"},{"task_slug":"physics-mc","task_name":"Physics MC"},{"task_slug":"presuppositions-as-nli","task_name":"Presuppositions As NLI"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"question-selection","task_name":"Question Selection"},{"task_slug":"reasoning-about-colored-objects","task_name":"Reasoning About Colored Objects"},{"task_slug":"riddle-sense","task_name":"Riddle Sense"},{"task_slug":"ruin-names","task_name":"Ruin Names"},{"task_slug":"snarks","task_name":"SNARKS"},{"task_slug":"sarcasm-detection","task_name":"Sarcasm Detection"},{"task_slug":"sentence-ambiguity","task_name":"Sentence Ambiguity"},{"task_slug":"sentence-completion","task_name":"Sentence Completion"},{"task_slug":"similarities-abstraction","task_name":"Similarities Abstraction"},{"task_slug":"sports-understanding","task_name":"Sports Understanding"},{"task_slug":"strategyqa","task_name":"StrategyQA"},{"task_slug":"temporal-sequences","task_name":"Temporal Sequences"},{"task_slug":"timedial","task_name":"Timedial"},{"task_slug":"understanding-fables","task_name":"Understanding Fables"},{"task_slug":"winowhy","task_name":"Winowhy"},{"task_slug":"word-sense-disambiguation","task_name":"Word Sense Disambiguation"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"attention-dropout","method_name":"Attention Dropout"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"chinchilla","method_name":"Chinchilla"},{"method_slug":"cosine-annealing","method_name":"Cosine Annealing"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gpt-3","method_name":"GPT-3"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"linear-warmup-with-cosine-annealing","method_name":"Linear Warmup With Cosine Annealing"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/analogical-similarity-on-big-bench","task":"Analogical Similarity","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"38.1"},"uses_additional_data":false},{"leaderboard":"/sota/analytic-entailment-on-big-bench","task":"Analytic Entailment","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"67.1"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-causal","task":"Common Sense Reasoning","dataset":"BIG-bench (Causal Judgment)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":4,"of":9,"metrics":{"Accuracy":"57.4"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-date","task":"Common Sense Reasoning","dataset":"BIG-bench (Date Understanding)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":5,"of":9,"metrics":{"Accuracy":"52.3"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench","task":"Common Sense Reasoning","dataset":"BIG-bench (Disambiguation QA)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":4,"of":9,"metrics":{"Accuracy":"54.7"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-known","task":"Common Sense Reasoning","dataset":"BIG-bench (Known Unknowns)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":3,"metrics":{"Accuracy":"65.2"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-logical","task":"Common Sense Reasoning","dataset":"BIG-bench (Logical Sequence)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"64.1"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-sports","task":"Common Sense Reasoning","dataset":"BIG-bench (Sports Understanding)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":4,"of":8,"metrics":{"Accuracy":"71"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-big-bench-winowhy","task":"Common Sense Reasoning","dataset":"BIG-bench (Winowhy)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":4,"metrics":{"Accuracy":"62.5"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-winogrande","task":"Common Sense Reasoning","dataset":"WinoGrande","model":"Chinchilla 70B (0-shot)","rank_in_archive_order":28,"of":77,"metrics":{"Accuracy":"74.9"},"uses_additional_data":false},{"leaderboard":"/sota/crash-blossom-on-big-bench","task":"Crash Blossom","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy ":"47.6"},"uses_additional_data":false},{"leaderboard":"/sota/crass-ai-on-big-bench","task":"Crass AI","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":3,"of":4,"metrics":{"Accuracy":"75.0"},"uses_additional_data":false},{"leaderboard":"/sota/dark-humor-detection-on-big-bench","task":"Dark Humor Detection","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy":"66.2"},"uses_additional_data":false},{"leaderboard":"/sota/discourse-marker-prediction-on-big-bench","task":"Discourse Marker Prediction","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"13.1"},"uses_additional_data":false},{"leaderboard":"/sota/empirical-judgments-on-big-bench","task":"Empirical Judgments","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"67.7"},"uses_additional_data":false},{"leaderboard":"/sota/english-proverbs-on-big-bench","task":"English Proverbs","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"82.4"},"uses_additional_data":false},{"leaderboard":"/sota/entailed-polarity-on-big-bench","task":"Entailed Polarity","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"94"},"uses_additional_data":false},{"leaderboard":"/sota/epistemic-reasoning-on-big-bench","task":"Epistemic Reasoning","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"60.6"},"uses_additional_data":false},{"leaderboard":"/sota/evaluating-information-essentiality-on-big","task":"Evaluating Information Essentiality","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"17.6"},"uses_additional_data":false},{"leaderboard":"/sota/fantasy-reasoning-on-big-bench","task":"Fantasy Reasoning","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"69"},"uses_additional_data":false},{"leaderboard":"/sota/figure-of-speech-detection-on-big-bench","task":"Figure Of Speech Detection","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"63.3"},"uses_additional_data":false},{"leaderboard":"/sota/gre-reading-comprehension-on-big-bench","task":"GRE Reading Comprehension","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"53.1"},"uses_additional_data":false},{"leaderboard":"/sota/general-knowledge-on-big-bench","task":"General Knowledge","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"94.3"},"uses_additional_data":false},{"leaderboard":"/sota/human-organs-senses-multiple-choice-on-big","task":"Human Organs Senses Multiple Choice","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"85.7"},"uses_additional_data":false},{"leaderboard":"/sota/identify-odd-metapor-on-big-bench","task":"Identify Odd Metapor","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"68.8"},"uses_additional_data":false},{"leaderboard":"/sota/implicatures-on-big-bench","task":"Implicatures","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"75"},"uses_additional_data":false},{"leaderboard":"/sota/implicit-relations-on-big-bench","task":"Implicit Relations","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"49.4"},"uses_additional_data":false},{"leaderboard":"/sota/intent-recognition-on-big-bench","task":"Intent Recognition","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"92.8"},"uses_additional_data":false},{"leaderboard":"/sota/irony-identification-on-big-bench","task":"Irony Identification","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"73.0"},"uses_additional_data":false},{"leaderboard":"/sota/lambada-on-big-bench","task":"LAMBADA","dataset":"BIG-bench","model":"Chinchilla-70B (zero-shot)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"77.4"},"uses_additional_data":false},{"leaderboard":"/sota/language-modelling-on-lambada","task":"Language Modelling","dataset":"LAMBADA","model":"Chinchilla (Zero-Shot)","rank_in_archive_order":16,"of":37,"metrics":{"Accuracy":"77.7"},"uses_additional_data":false},{"leaderboard":"/sota/logical-args-on-big-bench","task":"Logical Args","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy ":"56.2"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-formal","task":"Logical Reasoning","dataset":"BIG-bench (Formal Fallacies Syllogisms Negation)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":7,"of":9,"metrics":{"Accuracy":"52.1"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-logic-grid","task":"Logical Reasoning","dataset":"BIG-bench (Logic Grid Puzzle)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":4,"metrics":{"Accuracy":"44"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-logical","task":"Logical Reasoning","dataset":"BIG-bench (Logical Fallacy Detection)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"72.1"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-penguins-in-a","task":"Logical Reasoning","dataset":"BIG-bench (Penguins In A Table)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":3,"of":9,"metrics":{"Accuracy":"48.7"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-reasoning","task":"Logical Reasoning","dataset":"BIG-bench (Reasoning About Colored Objects)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":3,"of":9,"metrics":{"Accuracy":"59.7"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-strategyqa","task":"Logical Reasoning","dataset":"BIG-bench (StrategyQA)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":4,"metrics":{"Accuracy":"68.3"},"uses_additional_data":false},{"leaderboard":"/sota/logical-reasoning-on-big-bench-temporal","task":"Logical Reasoning","dataset":"BIG-bench (Temporal Sequences)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":5,"of":9,"metrics":{"Accuracy":"32.0"},"uses_additional_data":false},{"leaderboard":"/sota/mathematical-induction-on-big-bench","task":"Mathematical Induction","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy ":"47.3"},"uses_additional_data":false},{"leaderboard":"/sota/metaphor-boolean-on-big-bench","task":"Metaphor Boolean","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"93.1"},"uses_additional_data":false},{"leaderboard":"/sota/misconceptions-on-big-bench","task":"Misconceptions","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"65.3"},"uses_additional_data":false},{"leaderboard":"/sota/moral-permissibility-on-big-bench","task":"Moral Permissibility","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"57.3"},"uses_additional_data":false},{"leaderboard":"/sota/movie-dialog-same-or-different-on-big-bench","task":"Movie Dialog Same Or Different","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"54.5"},"uses_additional_data":false},{"leaderboard":"/sota/multi-task-language-understanding-on-mmlu","task":"Multi-task Language Understanding","dataset":"MML","model":"chatgpt/gpt3.5(20B)","rank_in_archive_order":14,"of":44,"metrics":{"Average (%)":"67.5"},"uses_additional_data":false},{"leaderboard":"/sota/multiple-choice-question-answering-mcqa-on-27","task":"Multiple Choice Question Answering (MCQA)","dataset":"BIG-bench (Hyperbaton)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":8,"of":9,"metrics":{"Accuracy":"54.2"},"uses_additional_data":false},{"leaderboard":"/sota/multiple-choice-question-answering-mcqa-on-28","task":"Multiple Choice Question Answering (MCQA)","dataset":"BIG-bench (Movie Recommendation)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":8,"of":9,"metrics":{"Accuracy":"75.6"},"uses_additional_data":false},{"leaderboard":"/sota/multiple-choice-question-answering-mcqa-on-29","task":"Multiple Choice Question Answering (MCQA)","dataset":"BIG-bench (Navigate)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":4,"of":9,"metrics":{"Accuracy":"52.6"},"uses_additional_data":false},{"leaderboard":"/sota/multiple-choice-question-answering-mcqa-on-31","task":"Multiple Choice Question Answering (MCQA)","dataset":"BIG-bench (Novel Concepts)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":2,"of":4,"metrics":{"Accuracy":"65.6"},"uses_additional_data":false},{"leaderboard":"/sota/multiple-choice-question-answering-mcqa-on-30","task":"Multiple Choice Question Answering (MCQA)","dataset":"BIG-bench (Ruin Names)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":8,"of":9,"metrics":{"Accuracy":"47.1"},"uses_additional_data":false},{"leaderboard":"/sota/nonsense-words-grammar-on-big-bench","task":"Nonsense Words Grammar","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"78"},"uses_additional_data":false},{"leaderboard":"/sota/odd-one-out-on-big-bench","task":"Odd One Out","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"70.9"},"uses_additional_data":false},{"leaderboard":"/sota/phrase-relatedness-on-big-bench","task":"Phrase Relatedness","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"94"},"uses_additional_data":false},{"leaderboard":"/sota/physical-intuition-on-big-bench","task":"Physical Intuition","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"79"},"uses_additional_data":false},{"leaderboard":"/sota/presuppositions-as-nli-on-big-bench","task":"Presuppositions As NLI","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"49.9"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"Chinchilla 70B (0-shot)","rank_in_archive_order":21,"of":65,"metrics":{"Accuracy":"83.7"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-natural-questions","task":"Question Answering","dataset":"Natural Questions","model":"Chinchilla (few-shot, k=64)","rank_in_archive_order":28,"of":47,"metrics":{"EM":"35.5"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-piqa","task":"Question Answering","dataset":"PIQA","model":"Chinchilla 70B (0-shot)","rank_in_archive_order":26,"of":67,"metrics":{"Accuracy":"81.8"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-social-iqa","task":"Question Answering","dataset":"SIQA","model":"Chinchilla (zero-shot)","rank_in_archive_order":19,"of":24,"metrics":{"Accuracy":"51.3"},"uses_additional_data":false},{"leaderboard":"/sota/question-selection-on-big-bench","task":"Question Selection","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"52.6"},"uses_additional_data":false},{"leaderboard":"/sota/riddle-sense-on-big-bench","task":"Riddle Sense","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"85.7"},"uses_additional_data":false},{"leaderboard":"/sota/sarcasm-detection-on-big-bench-snarks","task":"Sarcasm Detection","dataset":"BIG-bench (SNARKS)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":7,"of":8,"metrics":{"Accuracy":"58.6"},"uses_additional_data":false},{"leaderboard":"/sota/sentence-ambiguity-on-big-bench","task":"Sentence Ambiguity","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"71.7"},"uses_additional_data":false},{"leaderboard":"/sota/sentence-completion-on-hellaswag","task":"Sentence Completion","dataset":"HellaSwag","model":"Chinchilla 70B (0-shot)","rank_in_archive_order":42,"of":89,"metrics":{"Accuracy":"80.8"},"uses_additional_data":false},{"leaderboard":"/sota/similarities-abstraction-on-big-bench","task":"Similarities Abstraction","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"87"},"uses_additional_data":false},{"leaderboard":"/sota/timedial-on-big-bench","task":"Timedial","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"68.8"},"uses_additional_data":false},{"leaderboard":"/sota/understanding-fables-on-big-bench","task":"Understanding Fables","dataset":"BIG-bench","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy ":"60.3"},"uses_additional_data":false},{"leaderboard":"/sota/word-sense-disambiguation-on-big-bench","task":"Word Sense Disambiguation","dataset":"BIG-bench (Anachronisms)","model":"Chinchilla-70B (few-shot, k=5)","rank_in_archive_order":1,"of":6,"metrics":{"Accuracy":"69.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2203.15556","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2203.15556"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/karpathy/llama2.c","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/nkluge-correa/teenytinyllama","reach":null}],"summary":{"ran_fixture":2,"ran":3,"ran_draft_wrong":1,"ran_honours":2,"unverified":3},"by_repo_kind":{"listed":{"samples":7,"ran":4,"repositories":2}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":4,"samples":[{"code_sha256_prefix":"6d2a08dcf3466514","entry":"repeat_kv","repo":null,"repo_kind":null,"path":null,"file_url":null,"link_basis":"identical_code_first_harvested_elsewhere","language":"python","status":"ran_fixture","verification_level":2,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":null,"inline_ok":false,"mcp_get_code":{"code_sha256":"6d2a08dcf3466514"}},{"code_sha256_prefix":"1bf497baaf536b2b","entry":"FeedForward","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1bf497baaf536b2b"}},{"code_sha256_prefix":"7312586d85e4354f","entry":"ModelArgs","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7312586d85e4354f"}},{"code_sha256_prefix":"e27b34a3f741d4de","entry":"RMSNorm","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e27b34a3f741d4de"}},{"code_sha256_prefix":"af440c67b16afe67","entry":"apply_rotary_emb","repo":null,"repo_kind":null,"path":null,"file_url":null,"link_basis":"identical_code_first_harvested_elsewhere","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":null,"inline_ok":false,"mcp_get_code":{"code_sha256":"af440c67b16afe67"}},{"code_sha256_prefix":"9293e052d06482a7","entry":"calculate_loss","repo":"nkluge-correa/teenytinyllama","repo_kind":"listed","path":"Utilities/chinchilla-estimation.py","file_url":"https://github.com/nkluge-correa/teenytinyllama/blob/HEAD/Utilities/chinchilla-estimation.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"9293e052d06482a7"}},{"code_sha256_prefix":"5f447bdd807ed3c0","entry":"precompute_freqs_cis","repo":null,"repo_kind":null,"path":null,"file_url":null,"link_basis":"identical_code_first_harvested_elsewhere","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":null,"inline_ok":false,"mcp_get_code":{"code_sha256":"5f447bdd807ed3c0"}},{"code_sha256_prefix":"e04ad9d4b02f50b6","entry":"reshape_for_broadcast","repo":null,"repo_kind":null,"path":null,"file_url":null,"link_basis":"identical_code_first_harvested_elsewhere","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":null,"inline_ok":false,"mcp_get_code":{"code_sha256":"e04ad9d4b02f50b6"}},{"code_sha256_prefix":"c20a31b2d7615a4b","entry":"Attention","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c20a31b2d7615a4b"}},{"code_sha256_prefix":"8b8a4640c5a93610","entry":"Transformer","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8b8a4640c5a93610"}},{"code_sha256_prefix":"46b07f9445b01dce","entry":"TransformerBlock","repo":"karpathy/llama2.c","repo_kind":"listed","path":"model.py","file_url":"https://github.com/karpathy/llama2.c/blob/HEAD/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"46b07f9445b01dce"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}